Gamasutra - Sponsored Feature: Looks Aren't Everything: Making Games Act Real

[In this Intel-sponsored thought piece, company veteran Roger Chandler looks at why trudging through the Uncanny Valley to create realistic interactions with virtual creatures may lead to upsides in AI, special effects, physics, and more.]

One night several months ago, my buddies and I logged on to play a newly released, online role-playing game. The environments were lush, the details were rich, and the monsters were stunningly rendered. I was quickly sucked into it all...until I ran into a small log lying in front of me. It literally stopped me in my tracks.

I know a lot of great game developers and designers, and I understand the difficult design decisions they must make when bringing a title to market, but this log really surprised me. It was richly textured and accurately modeled, which was nice, but it "behaved" like a brick wall.

Despite being a powerful warrior who had just slain three ogres single-handedly, I could not raise my in-game foot 18 virtual inches to pass over this small log. So I walked around it and we continued on. But that experience stuck with me.

In 1970 roboticist Masahiro Mori introduced the Uncanny Valley hypothesis: as robots and other representations of humans begin to look and act almost, but not entirely, like actual humans, it causes a response of revulsion among human observers.

The "valley" refers to a precipitous drop in the viewers' positive response to the near-human entity as it gets closer to realism, due to the dramatically increased expectations around behavior and other subtle human-like details.

The Uncanny Valley has drawn much attention from the gaming industry in recent years, and while it definitely applies to human-like characters, I think the effect extends to the environment as a whole.

As a first-hand witness to our industry's ongoing graphics arms race, I believe it takes more than beautiful scenery to engage a gamer. We live in a world where every object has distinct physical properties and every creature has unique behavioral characteristics.

As developers continue to improve how these objects and creatures look in-game, they must also meet the players' heightened expectations regarding how these objects and creatures will act.

For example, I do not expect a blocky, pixelated tree to sway in the wind or splinter realistically when I blow it to bits with a rocket launcher. But if that tree looks nearly identical to the one in my front yard, then it will be a noticeable distraction if it does not act like the real thing.

One of the things I love about my job is hearing developers' thoughts on how they can create new levels of interactive realism with the increasingly powerful and programmable multi-core processors Intel has on its roadmap.

Certainly developers are doing great things with today's generation of multi-core processors to ensure games look and act more real. Here are a few of their efforts:

Comments

Andrew Heywood

18 Jul 2008 at 7:36 am PST

o rly

Lou Hayt

19 Jul 2008 at 8:37 pm PST

This is a great article, I think it touches a fundamental issue in game design - lets take a closer look at the tavern brawl example: one side of the coin could be that a brawl at the tavern is a carefully designed mini game that may affect the game in a specific, carefully designed way. The other side of the coin (as I understand it) is that the tavern and it patrons were simulated in high fidelity, therefore smashing a chair on my friend's face is a possible action within the tavern's simulated environment. This will probably cause a chain reaction that form "emergent behavior". It is impossible to predict this behavior, it is only possible to iterate on the design. After a big brawl... who is going to clean the tavern? what if all the chairs are smashed to pieces and there is no where to sit?

What are the pros and cons of these two approaches? A carefully designed interaction is easier to measure to see if we achieved our goals, the goals of a simulated environment are inheritely more complex... and more player driven. On the other hand we only need to write a breakable chair system once and it applies wherever we place a breakable chair, it may apply to more situations than a brawl - a player might be able to stack those chairs and climb to the roof, use it to fuel the fireplace, or who knows for what - if the simulation enables the player with self expression within loose boundaries.

Tim Elder

20 Jul 2008 at 6:53 pm PST

Nice article. It's exciting to think that we might one day have the processing power to be able to do all of what you mention, but my question is this - What benefit to the gameplay experience does having "individually rendered strands of hair" give?

Not matter the processing power, funding, or time available, developers will always have finite resources available to create these products, and finite CPU cycles to run them in and still keep the frame rate up. The difference comes down to which battles we choose to fight. If you choose to focus on one area, then something else will ultimately have to give.

Look at Dwarf Fortress - the only way the developer could have the depth of possible interactions and immersive dynamic player-driven story is by drawing a line in the sand with the graphics and saying "sprite sheets are enough".

Near the other end of the spectrum, Crysis is a stunningly beautiful multimedia experience, but the gameplay is little more detailed than any other dynamic physics enabled FPS. No offence to Crytek, it's a solid game and has some neat innovations, but the focus has clearly been on the graphics.

So, if you're going to have the "individually rendered strands of hair" feature, then what are you going to take out (or hold back) to make it fit into your development schedule?

Ian Wilson

21 Jul 2008 at 11:50 am PST

This sounds like just the area a company like Intel would want to put their marketing dollars. These innovations are here now, the future is here, it is just unevenly distributed (to mangle a well known saying).

What developers need to see are examples of these technologies working together, like movie quality characters with deeply emotional real time, interactive behaviors. But what the developers of these technologies would benefit from is the support of someone like Intel to help get those (CPU hungry) technologies out in front of the market.

Sounds like a win-win.

Sebastian Bender

22 Jul 2008 at 1:40 pm PST

Since Tim referred to Crysis: I had a similar experience like Mr. Chandler had in AoC in Crysis. I saw this stunningly beautiful environment and I really enjoyed the physics and exploring the possibilities (the gameplay itself was "not so felicitous"). But while playing around I occasionally ran into a bush - my biggest mistake. Looking at that bush I really couldn't believe my eyes. It was simply an animated plane with a quite ugly texture on it. I stumbled down the Uncanny Valley. So I guess that "individually rendered strands of hair" can really help delivering a message to the player or at least help avoiding such immersion-breakers.
I get the point that creating these things takes extra time but as long as there are games that profile themselves through graphics this time needs to be invested.
I still prefer gameplay over graphics though. And advanced technology may also be the key to offering new possibilities in gameplay.

Jeremy Hayes

25 Jul 2008 at 1:48 pm PST

"I stumbled down the Uncanny Valley." Great quote Sebastien. Love it.

I think Tim Elder made some really good points, and I couldn't agree more about Crysis--great graphics, but lackluster gameplay IMO.

I've been very disappointed by state-of-the-art game AI. I think we're a long way from the type of "intelligent ogres" Mr. Chandler is referring to. Of course, I hope I'm wrong.